Annotation Data Analysis

# Importing libraries
import yaml
import pymongo
import pandas as pd
from collections import Counter

# Import plot libraries
import plotly.express as px

Util functions

# File function - Read dict from YAML file
def get_dict_from_yaml(yaml_path:str, encoding:str="utf-8") -> dict:
    result = dict()
    
    try:
        with open(yaml_path, mode="r", encoding=encoding) as file:
            yaml_file = file.read()
            result = yaml.load(yaml_file, Loader=yaml.FullLoader)
    except Exception as e:
        print(e)
        
    return result

def string_to_list(str_list:str) -> list:
    if str_list == "[]":
        return []
    str_list = str_list.replace("[", "").replace("]", "").replace(" ", "").replace("\n", "")
    return str_list.split(",")

# Util function - Plot column chart
def plot_bar_chart(df, x_var, y_var, title, color=None):
    fig = px.bar(df, x=x_var, y=y_var, orientation='h', title=title, color=color)
    fig.update_layout(yaxis={'categoryorder':'total ascending'}, showlegend=True)
    fig.show()

Reading data

# Read MongoDB login info
filepath = "config/mdb_setup.yaml"
mdb_login = get_dict_from_yaml(filepath)

# Create connetion to MongoDB
client  = pymongo.MongoClient(mdb_login['db_server'], mdb_login['db_port'])
db = client[mdb_login['db_name']]
coll = db[mdb_login['db_collection']]

# Query data
doc_list = []
for doc in coll.find({}):
    doc['claim']['nouns'] = string_to_list(doc['claim']['nouns'])
    doc['claim']['entities'] = string_to_list(doc['claim']['entities'])
    doc['majorClaim']['nouns'] = string_to_list(doc['majorClaim']['nouns'])
    doc['majorClaim']['entities'] = string_to_list(doc['majorClaim']['entities'])
    doc['premise']['nouns'] = string_to_list(doc['premise']['nouns'])
    doc['premise']['entities'] = string_to_list(doc['premise']['entities'])
    doc_list.append(doc)

print(">> Number of annotated documents:", len(doc_list))

>> Number of annotated documents: 4702

doc_list[0]

{'_id': ObjectId('618315e580c699e1073e9a16'),
 'argumentID': '5-1',
 'approach': 'A2 -> C+(L+P)',
 'claim': {'text': 'Mi idea es dar un uso al arco de la victoria de Moncloa , en lugar de demolerlo , como soporte',
  'nouns': ['idea',
   'uso',
   'arco',
   'victoria',
   'lugar',
   'demolerlo',
   'soporte'],
  'entities': ['Moncloa']},
 'linker': {'linker': 'para',
  'category': 'CONSEQUENCE',
  'subCategory': 'GOAL',
  'relationType': 'support'},
 'mainVerb': 'dar',
 'majorClaim': {'text': 'Uso artistico y de promocion del  Arco de la Victoria',
  'nouns': ['Uso', 'promocion'],
  'entities': ['Arco', 'laVictoria']},
 'premise': {'text': 'promos de actividades en Madrid',
  'nouns': ['promos', 'actividades'],
  'entities': ['Madrid']},
 'proposalID': 5,
 'sentence': 'Mi idea es dar un uso al arco de la victoria de Moncloa, en lugar de demolerlo, como soporte para promos de actividades en Madrid (semana del orgullo gay, congresos mundiales, eventos futbol, etc)'}

Data Processing

# Local variables
n_top = 20
topic_filter = ""
ls_linkers = []
entities = Counter()
nouns = Counter()
main_verbs = Counter()

# Counting data
for doc in doc_list:
    if doc["sentence"] == "" or topic_filter in doc["sentence"]:
        
        # Linker data
        doc_linker = doc["linker"]
        ls_linkers.append([doc_linker["category"], doc_linker["subCategory"], doc_linker["linker"], doc_linker["relationType"]])
        
        # Entities data
        entities_set = set(doc["claim"]["entities"] + doc["majorClaim"]["entities"] + doc["premise"]["entities"])
        for entity in entities_set:
            entities[entity] += 1
            
        # Nouns data
        nouns_set = set(doc["claim"]["nouns"] + doc["majorClaim"]["nouns"] + doc["premise"]["nouns"])
        for entity in nouns_set:
            nouns[entity] += 1
            
        # Verbs data
        if doc["mainVerb"] != None:
            main_verbs[doc["mainVerb"].lower()] += 1

# Cooking linkers dataframe
df_linkers = pd.DataFrame(ls_linkers, columns=["category", "subCategory", "linker", "relationType"])
df_linkers.head(10)

	category	subCategory	linker	relationType
0	CONSEQUENCE	GOAL	para	support
1	CONSEQUENCE	GOAL	para	support
2	EXPLANATION	EXEMPLIFICATION	como por ejemplo	support
3	CAUSE	CONDITION	si	qualifier
4	CONSEQUENCE	GOAL	para	support
5	CONSEQUENCE	GOAL	para	support
6	CONSEQUENCE	GOAL	para	support
7	CONSEQUENCE	GOAL	para	support
8	CONSEQUENCE	GOAL	para	support
9	CAUSE	REASON	porque	support

1. Most used categories

# Cooking dataframe
df = df_linkers.groupby(['category']).size().reset_index(name='counts')
df.rename(columns={'counts': 'frequency'}, inplace=True)
df.sort_values(by='frequency', ascending=False, inplace=True)
df

	category	frequency
1	CONSEQUENCE	3226
0	CAUSE	1189
4	EXPLANATION	167
2	CONTRAST	119
3	ELABORATION	1

# Plot top N most used categories
x_var = 'frequency'
y_var = 'category'
title = 'Top %s Most used categories' % min(n_top, len(df))
plot_bar_chart(df, x_var, y_var, title)

2. Most common sub-categories

# Cooking dataframe
df = df_linkers.groupby(['category', 'subCategory']).size().reset_index(name='counts')
df.rename(columns={'counts': 'frequency'}, inplace=True)
df.sort_values(by='frequency', ascending=False, inplace=True)
df.head(10)

	category	subCategory	frequency
2	CONSEQUENCE	GOAL	3158
0	CAUSE	CONDITION	662
1	CAUSE	REASON	527
6	EXPLANATION	EXEMPLIFICATION	137
4	CONTRAST	ALTERNATIVE	119
3	CONSEQUENCE	RESULT	68
7	EXPLANATION	RESTATEMENT	29
5	ELABORATION	ADDITION	1
8	EXPLANATION	SUMMARY	1

# Plot top N most used sub-categories
x_var = 'frequency'
y_var = 'subCategory'
title = 'Top %s Most used sub-categories' % min(n_top, len(df))
color_by = 'category'
plot_bar_chart(df, x_var, y_var, title, color_by)

3. Most used linkers by category

# Cooking dataframe
df = df_linkers.groupby(['category', 'linker']).size().reset_index(name='counts')
df.rename(columns={'counts': 'frequency'}, inplace=True)
df.sort_values(by='frequency', ascending=False, inplace=True)
df.head(10)

	category	linker	frequency
56	CONSEQUENCE	para	3115
32	CAUSE	si	445
35	CAUSE	ya que	228
29	CAUSE	porque	160
77	EXPLANATION	por ejemplo	90
3	CAUSE	ante	67
30	CAUSE	pues	49
67	CONTRAST	ni	42
11	CAUSE	debido a	40
68	CONTRAST	o	37

# Plot top N most used linkers
x_var = 'frequency'
y_var = 'linker'
title = 'Top %s Most used linkers' % min(n_top, len(df))
color_by = 'category'
plot_bar_chart(df.head(n_top), x_var, y_var, title, color_by)

4. Most used linkers by relationType

# Cooking dataframe
df = df_linkers.groupby(['relationType', 'linker']).size().reset_index(name='counts')
df.rename(columns={'counts': 'frequency'}, inplace=True)
df.sort_values(by='frequency', ascending=False, inplace=True)
df.head(10)

	relationType	linker	frequency
59	support	para	3115
16	qualifier	si	445
73	support	ya que	228
68	support	porque	160
60	support	por ejemplo	90
2	qualifier	ante	67
69	support	pues	49
80	support/attack	ni	42
46	support	debido a	40
13	qualifier	frente a	37

# Plot top N most used linkers
x_var = 'frequency'
y_var = 'linker'
title = 'Top %s Most used linkers' % min(n_top, len(df))
color_by = 'relationType'
plot_bar_chart(df.head(n_top), x_var, y_var, title, color_by)

5. Most common entities

# Cooking dataframe
top_entities = entities.most_common(n_top)
df = pd.DataFrame.from_records(top_entities, columns = ['entities', 'frequency'])
df["% share"] = round(100.0 * df["frequency"] / sum(entities.values()), 2)
df.head(10)

	entities	frequency	% share
0	Madrid	538	8.82
1	Ayuntamiento	102	1.67
2	EMT	58	0.95
3	Vallecas	42	0.69
4	MADRID	40	0.66
5	Retiro	39	0.64
6	Metro	38	0.62
7	Rio	36	0.59
8	AyuntamientodeMadrid	32	0.52
9	M-30	31	0.51

# Plot top N most used entities
x_var = 'frequency'
y_var = 'entities'
title = 'Top %s Most used entities' % min(n_top, len(df))
plot_bar_chart(df.head(n_top), x_var, y_var, title)

6. Most common nouns

# Cooking dataframe
top_nouns = nouns.most_common(n_top)
df = pd.DataFrame.from_records(top_nouns, columns = ['nouns', 'frequency'])
df["% share"] = round(100.0 * df["frequency"] / sum(nouns.values()), 2)
df.head(10)

	nouns	frequency	% share
0	calle	367	1.36
1	zona	336	1.24
2	calles	259	0.96
3	barrio	258	0.95
4	personas	216	0.80
5	parque	212	0.78
6	ciudad	210	0.78
7	uso	189	0.70
8	transporte	189	0.70
9	zonas	186	0.69

# Plot top N most used nouns
x_var = 'frequency'
y_var = 'nouns'
title = 'Top %s Most used nouns' % min(n_top, len(df))
plot_bar_chart(df, x_var, y_var, title)

7. Most common main verbs

# Cooking dataframe
top_main_verbs = main_verbs.most_common(n_top)
df = pd.DataFrame.from_records(top_main_verbs, columns = ['main verbs', 'frequency'])
df["% share"] = round(100.0 * df["frequency"] / sum(main_verbs.values()), 2)
df.head(10)

	main verbs	frequency	% share
0	propongo	184	4.23
1	crear	171	3.94
2	poner	82	1.89
3	hacer	66	1.52
4	ampliar	62	1.43
5	creo	54	1.24
6	tiene	52	1.20
7	gustaria	48	1.10
8	habilitar	47	1.08
9	tienen	46	1.06

# Plot top N most used main verbs
x_var = 'frequency'
y_var = 'main verbs'
title = 'Top %s Most used main verbs' % min(n_top, len(df))
plot_bar_chart(df, x_var, y_var, title)

---

« Back to the repository.